Search pathology Games and search Bioinformatics Tourism
Machine learning Genre classification Other

Search pathology

Search pathology is a phenomenon best known from minimax search, where under seemingly reasonable conditions, the deeper one searches, the worse he plays - the opposite of what happens in practice. Similar behavior was observed in real-time single-agent heuristic search. Search pathology was the subject of my Ph. D. thesis under the supervision of Ivan Bratko and Matjaž Gams. My research on the pathology in single-agent search was done in collaboration with Vadim Bulitko.

Publications

Other documents

Bioinformatics

I started working in this area as a postdoc at the Institute for Biostatistics and Informatics in Medicine and Ageing Research. I developed a machine-learning method for epitope prediction based on peptide array data, which is relevant to vaccine design and diagnostics. I was also involved in the research on pluripotency.

I am currently involved in the COVIRNA project, which aims develop a diagniostic panel to predict outcomes of cardiovascualr patients with COVID-19.

Publications

Other documents

Tourism

We have developed an intelligent electronic tourist guide as a web and mobile application. The guide prepares a personalized itinerary for each user and then guides him on his trip. The personalization relies on knowledge-based recommendations and collaborative filtering.

Publications

Games and search

Computer game playing was the subject of my B. Sc. thesis, for which I wrote a program for playing tarok called Silicon Tarokist. The program was substantially improved afterwards. I later also collaborated with Domen Marinčič, who worked on a Bayesian decision model for bidding in tarok. I have also done some research on real-time single-agent heuristic search, where I was investigating methods to determine the optimal lookahead depth. This research is an offshoot of the research on search pathology and was also done in collaboration with Vadim Bulitko.

Publications

Other documents

Machine learning

While I use machine learning in most of my research, I do not often research machine learning itself. My main interest in this area is building classifiers that are both comprehensible and accurate.

Publications

Genre classification

The ability to classify web pages into genres can be a helpful addition to a search engine. We developed machine learning methods for genre classification in Alvis EU FP6 project, but the development continues after the end of the project. I collaborate on this with Vedrana Vidulin.

The web genre dataset used in our research

Publications

Other documents

Other

Occasionally I also dabble in other areas.

Publications